FEATURE
Issue: 

Finishing school for optics

Jessica Rowbury looks at the advances being made in optical polishing techniques

Polishing is one of the most important stages of transforming optical material into a finished component capable of controlling light in the desired manner. The grinding and polishing technique will depend largely on what the optic  will be used for, although to a certain extent advanced polishing capabilities are useless without the ability to test the finished product. Therefore, a lot of the progress being made in optical manufacture concerns metrology, as new polishing processes cannot be developed without the equipment to test them.

All optical components start out as raw material, which is first cut to the length, width and diameter required, but is often left thicker to allow for subsequent grinding and polishing. The raw material can vary, but, ‘generally, in our business, 80 to 85 per cent is N-BK7 or fused silica, and the other 15 to 20 per cent is made up of the more exotic materials, such as calcium fluoride or magnesium fluoride or specialist optical glasses,’ said Dr Helmut Kessler, managing director of Manx Precision Optics (MPO).

Before polishing takes place, the part is ground to reach an even thickness. A surface grinder removes any machining marks, followed by further grinding with a different sized grid – this gives the optics a uniform grey colour.

‘Often, the importance of the grinding process is underestimated,’ Kessler explained. ‘What is important is the grid size you use, and also the pressure you exert on the substrate during the grinding process. It is important simply because if you use the wrong grid or the wrong pressure, then you cause sub-surface damage in the raw material.’

After the raw material is cut to size and ground, it is now ready for polishing. In general, there are two techniques used to polish optics. Double sided polishing is used for parallel parts, such as square, rectangular, trapezoidal, round, elliptical or polygonal optics. The process uses a planetary action, with the combination of a top and bottom plate. The part is placed in between two polishing pads, primarily polyurethane-based with additives, and, in combination with a polishing slurry containing abrasive grains, the optic is polished simultaneously on the up-side and the down-side.

Pitch polishing is used for single sided optics, so is the technique of choice for non-parallel components such as prisms, spherical and cylindrical lenses, as well as components with ultra-high surface flatness. The optic is placed on a pitch lap – a black, extremely viscous liquid – which is used as the polishing pad. The pitch varies in hardness depending on the material being polished, so, for example, a softer material such as calcium fluoride would require a softer pitch than a harder material like BK7, Kessler noted.

Every machining operation leaves machining marks in the form of finely spaced micro irregularities. The pattern left on the surface is known as surface roughness, which, through polishing, is reduced to a smooth finish.

‘If a surface is rough then light can be diffracted, or a coating might not stick very well, or it might be difficult to clean,’ commented Dr Dirk Apitz, application engineer at Schott. ‘For most applications you have to have very smooth surfaces below a certain roughness, which is also specified together with the geometrical parameters of the component.’

It is possible to achieve better surface roughness, or RMS, using pitch polishing than with double sided polishing, according to Kessler. ‘The real high quality finish is achieved through pitch polishing… This is really important for the high power optics, to have a very good roughness. You can achieve one to two angstrom RMS on just straight forward pitch polishing, and this is pretty good roughness – you would see an atom lying on the surface,’ Kessler said.

However, according to Schott’s Apitz, although the flatness of the surface is more difficult to control through double sided polishing, the process can be used to achieve a very good transmitted wavefront distortion (a measure of the distortion a plane wave of light undergoes when transmitted through an optical element).

Transmitted wavefront distortion is particularly important for components such as windows, laser plates, and laser rods, where the purpose of the part is to transmit the wavefront without changing its shape.

Optics in use

Just as the polishing processes vary, the parameters required for a component will vary depending on its intended use.

‘If the component is a high-end mirror, then what is important is the reflected wavefront distortion and therefore the flatness of the surface. But with a transmissive component such as a window… what is important might not be the flatness, but the total thickness variation – so it must be equally thick over every part of the window. And so the parameter that gets specified usually is the transmitted wavefront distortion,’ Apitz remarked.

For high-power laser applications, the scratch-dig specification is important. This represents the maximum allowable size and number of scratches (a mark or tear) and digs (a small rough spot or divot) allowed on the surface – the tighter the scratch-dig specification, the lower the scatter. ‘You don’t want scratches or digs because they influence the growth of the coating, and therefore lower the laser-induced damage threshold,’ noted Kessler.

‘As an example, some of the optics we make for ultrafast lasers can withstand 0.4J/cm2 and 30 femtoseconds and 750-850nm – that is 500 to 1,000 times the power in the national grid. That is an enormous power level,’ he added.

However, with so many aspects to take into consideration, it is important not to over specify an optic. Kessler commented: ‘It is very easy to over specify an optic, and therefore needlessly spend money. It is very important to sit the customer down to find out what they actually want to do with the optic and only then you can find what they really need.’

Driving innovation

In order to stay competitive, processes continuously need to be developed and improved as specifications from customers become tighter and tighter. According to Apitz, much of this comes as a result of taking on specifications that are particularly challenging, as this means processes are pushed to the limit. ‘We know that we have to invest, we have to become better, we have to listen to our customers, and not just listen to what they want right now, but also what will be needed in a few years’ time.

‘There is a lot of customer interaction in that respect – [the customer] comes up with something that might be nice to have in three or four years’ time, and then they talk with us about whether that is feasible, or what they can work with instead,’ Apitz said. ‘I remember there were some small laser rods… the first time I first saw the specification I thought that no one on the planet could do this. And so there were two things we were doing – on the one hand we were pushing our process, and on the other hand, we discussed with the customer what we think is possible at reasonable costs.’

If you can’t test it, you can’t make it

Developing a new manufacturing process comes to a large extent from being able to measure and test that process. Kessler noted: ‘The improvements in polishing have only been made possible because of the improvements in metrology. The more accurately you can measure, the more easily you can develop a process. So, one really feeds off the other.’

Not only have metrology instruments dropped in price in the last few years, but their performance has also improved dramatically. ‘To test the absolute thickness we have got a laser interferometry gage which can measure to within fractions of a micron accuracy,’ Kessler said. ‘Even when you look back 10-15 years, there were certain aspects in interferometry that people could not really measure reliably, and now, it’s standard.

‘If you can’t test it, you can’t make it,’ Kessler added. ‘When we [MPO] first started up two and a half years ago we looked into purchasing second-hand metrology, and then discovered it wasn’t feasible because second hand or inferior metrology does not allow you to measure some of the parameters and specifications which are required nowadays,’ Kessler remarked. ‘If you want to be a leading manufacturer, you have to have leading metrology.’

Schott has recently purchased a magnetic rheological finishing (MRF) machine in order to polish aspheric lenses with shape deviations down to a few tens of nanometres. Apitz said: ‘It’s [MRF] a local correction of the surface – you polish the surface, then you measure the deviation between the shape that the lens has and what it should be, and then you can calculate on which spot on the surface you still have to remove some material.’                                                           This process can also be applied to laser rods, for correcting the transmitted wavelength distortion more accurately. ‘If you want the best transmitted wavelength distortion, having perfectly flat surfaces might not be the key factor, as possible refractive variations in the material will determine transmitted wavelength distortion,’ explained Apitz. ‘But you can measure the distortion and then locally inversely correct one surface by using an MRF process.’

‘With this technology we will be able to compensate these influences in long and thick laser glass rods to reach a transmitted wavefront  distortion in the range of lambda/20,’ Apitz added.’

Working with broadband light

As customers are working more with visible light in addition to the more traditional infrared, optics are increasingly required to perform well over a broader range of wavelengths. And, this requires looking at how the mid-spatial frequencies affect the reflection of light in addition to the Ra surface roughness values and scratch-dig specifications.

‘A lot of applications are becoming more and more broadband – so customers are not just using lasers, but white light sources, supercontinuum lasers, visible diodes,’ noted Mark Wilkinson, director of Laser Beam Products. ‘The days of having a mirror specified just for the infrared are becoming numbered; people want a multifunctional mirror that will work not just in the IR but perhaps for the visible as well.’

In addition, although the final application of the mirror might be for an infrared application such as CO2 laser cutting, where surface roughness values in the range of tens of nanometres are often adequate, it may not be possible to test or align the part because these techniques often involve the use of visible light. ‘You might be using terahertz or CO2 lasers – all these long wavelength applications where the quality of the mirrors doesn’t matter too much. But, you have to align it or test it, and that generally requires some sort of visible laser or visible technique,’ Wilkinson explained. ‘Then there will be scatter and diffraction of the visible light, and although this won’t matter when it is installed in the customer’s premises, if you can’t line it up in the factory then you are a bit stuck.

‘This is something that people overlook, we’ve had customers who think the mirrors don’t need to be that high quality for their terahertz application, but then when they try to align their laser they cannot do it,’ Wilkinson added.

Metal mirrors used for CO2 and terahertz applications are often produced by Single Point Diamond Turning (SPDT), whereby a flat, spherical, aspheric or even a freeform reflective surface is machined directly onto the mirror.

However, a simple surface roughness value or a scratch-dig specification has often been found to be inadequate for SPDT metal mirrors used with visible or near infrared radiation. Using white light interferometric testing, Laser Beam Products produced a cross section profile of a typical SPDT surface. Several ‘families’ of grooves could be observed with different spacings and amplitudes; this significantly reduces the amount of specularly reflected light.

So, although the SPDT parabolic mirror had a good surface roughness value of Ra = 5nm, when the grooves were analysed by their spatial frequency, a regularly repeating set of grooves could be seen with a spatial frequency in the order of 50 to 100 lines per millimetre. This repetitive ‘mid spatial frequency surface roughness’ rendered the mirror unusable at 1µm wavelength, and produced a large amount of scattered and diffracted light at 633nm.

As a result of polishing, the surface roughness spatial frequencies of the same mirror were reduced considerably.

‘If you want good quality reflection from a mirror surface, then surface roughness Ra doesn’t tell the whole story,’ Wilkinson said. ‘It’s the roughness and the important mid-spatial frequencies content which control the qualities you get. So, simply having a surface roughness of Ra = 5nm is just not sufficient.’